Skip to main content

Secret scrubbing

Before upload, skl scans every file in the skill tree. Findings are block or warn. Blocks always stop the upload. Warns stop the upload unless you pass --allow-warnings.

skl sync --allow-warnings

--allow-warnings does not override a block.

Blocks​

These refuse upload:

  • Private keys (BEGIN … PRIVATE KEY)
  • AWS access keys and aws_secret_access_key assignments
  • GitHub tokens (ghp_…, github_pat_…)
  • Slack tokens (xoxb-… and related)
  • Stripe live keys (sk_live_…)
  • OpenAI and Anthropic keys
  • npm tokens (npm_…)
  • Env-style assignments for api_key, secret_key, access_token, auth_token, password, or private_key

.env, .env.local, .env.production, and .env.development are scanned with the rest of the tree.

Warns​

These need --allow-warnings to upload:

  • JWTs
  • Google API keys (AIza…)

Still check the skill​

The scanner looks for obvious credential patterns. It will miss things. Do not put secrets in SKILL.md or in files next to it.